Loops were firstly called using the ABC Model. Since this method integrates ATAC-seq, H3K27ac and gene expression data when calling enhancer-promoter (E-P) pairs, we can be more confident we are truly calling loops. After defining E-P loops, differential loop calling was performed using a modified DESeq2 approach via Hictoolsr.

ABC model results

The quantitative effect of an enhancer depends on the frequency with which it contacts a promoter multiplied by the strength of the enhancer (i.e., the ability of the enhancer to activate transcription upon contacting a promoter). Moreover, the contribution of a specific enhancer to a gene’s expression should depend on the surrounding context (ie, the strength and contact frequency of other enhancers for the gene). To convert this conceptual framework into a practical score (which can be applied genome-wide), the ABC score was formulated:

Activity (A) is calculated as the geometric mean of the read counts of ATAC-seq and H3K27ac chromatin immunoprecipitation sequencing (ChIP–seq) at element E, and Contact (C) as the KR-normalized Hi-C contact frequency between E and the promoter of gene G at 5-kb resolution.

Download Enhancer_promoter_loops.csv

Differential loop analysis

Differential loop analysis was performed at 1000kb resolution (ie. regions were binned at 10kb windows).

Raw counts

Counts

with chrX

without chrX

E-P pairs

DESeq2 res

with chrX

without chrX


PCA

Top 500 most variable loops

with chrX

without chrX

All loops

with chrX

without chrX

MA plot

with chrX

without chrX